Nice idea! I notice that this is one-level more general than Nuño's idea of paying for the initial liquidity of a personal AMM that you hand to a forecaster to profit from.
If you gave out multiple accuracy agreements, say to the first and second-place bidders (which both pay the second place bidder's price), then those parties would have good incentives to work together, sharing information and analysis. This incentive to cooperate is one way that an accuracy agreement could be a better structure than a prediction market.
Also, for forecasts that require a lot of upfront research, it can be more efficient to have only one or a couple parties pay the research cost. With a prediction market, you would need to collude with other parties so that only a max number do the costly research, which is a hard coordination problem. (Actually, according to your example, bidders submit their forecasts with the bid. Seems like because of research and analysis costs they should only submit forecasts after winning the bid?)
Thirdly, I really like the generic nature of having any computable scoring. I feel like there must be some powerful use cases here. If the purchaser trusts the client's judgment, the purchaser could submit qualitative predictions (e.g. prose) and the client could later give a score for how well they did.
An unsolved problem for prediction markets with an AMM is news traders taking more profit relative to effort expended. This accuracy agreement doesn't solve this problem. If a decisive event happens early that could earn them a lot of money, then they could rest and vest. In fact, because of the potential for profit, there could be some incentive for them to act to cause certain outcomes. (Just like with prediction markets!)
The legality problems are interesting, but you might also fare better than prediction markets. It doesn't fit into a known illegal category (AFAIK), and seems more likely to succeed at the argument that it's a game of skill. There also wouldn't be much legal risk until the agreements get much larger.
There's a lot here, I like these lines of thinking.
> This incentive to cooperate is one way that an accuracy agreement could be a better structure than a prediction market.
I could see this, though if actors can resell their contracts, then there might be some privacy. A potential buyer of a contract might underplay their hand, but actually, know of clever ways of reducing the uncertainty more than expected.
Meanwhile in prediction markets, there are of course actors who buy a big trade, then share selective information to support their trade. So there's at least some collaboration in each.
I'm most excited about collaboration via additional services of just estimating the value of collaboration and directly rewarding people, but that's more complicated.
>Actually, according to your example, bidders submit their forecasts with the bid. Seems like because of research and analysis costs they should only submit forecasts after winning the bid
I think "submitting forecasts with the bid" is simpler, for a thought experiment, but in practice mediocre, for this reason. I'd be more optimistic about submitting forecasts after, or being "continuously scored" using updating forecasts.
For example, a certain party could have been paid to forecast COVID cases, then be expected to forecast weekly - then later paid conditional on their accuracy.
> the purchaser could submit qualitative predictions (e.g. prose) and the client could later give a score for how well they did.
This is a bit of a different topic, but I've thought about this, and could be excited by it, especially if we could use AI. Like, we could hypothetically use AI to judge the accuracy of qualitative forecasts made by all historical figures and twitter intellectuals.
> An unsolved problem for prediction markets with an AMM is news traders taking more profit relative to effort expended. This accuracy agreement doesn't solve this problem.
I imagine the ease would be "priced in". If predictors thought that the first 3 bits of information are very easy to get, they'd submit an according bid for that.
>If a decisive event happens early that could earn them a lot of money, then they could rest and vest.
Sure, but they still could make even more money by doing better. They could also sell the contract/agreement, to someone interested in extracting more money from it.
>In fact, because of the potential for profit, there could be some incentive for them to act to cause certain outcomes.
Nice idea! I notice that this is one-level more general than Nuño's idea of paying for the initial liquidity of a personal AMM that you hand to a forecaster to profit from.
If you gave out multiple accuracy agreements, say to the first and second-place bidders (which both pay the second place bidder's price), then those parties would have good incentives to work together, sharing information and analysis. This incentive to cooperate is one way that an accuracy agreement could be a better structure than a prediction market.
Also, for forecasts that require a lot of upfront research, it can be more efficient to have only one or a couple parties pay the research cost. With a prediction market, you would need to collude with other parties so that only a max number do the costly research, which is a hard coordination problem. (Actually, according to your example, bidders submit their forecasts with the bid. Seems like because of research and analysis costs they should only submit forecasts after winning the bid?)
Thirdly, I really like the generic nature of having any computable scoring. I feel like there must be some powerful use cases here. If the purchaser trusts the client's judgment, the purchaser could submit qualitative predictions (e.g. prose) and the client could later give a score for how well they did.
An unsolved problem for prediction markets with an AMM is news traders taking more profit relative to effort expended. This accuracy agreement doesn't solve this problem. If a decisive event happens early that could earn them a lot of money, then they could rest and vest. In fact, because of the potential for profit, there could be some incentive for them to act to cause certain outcomes. (Just like with prediction markets!)
The legality problems are interesting, but you might also fare better than prediction markets. It doesn't fit into a known illegal category (AFAIK), and seems more likely to succeed at the argument that it's a game of skill. There also wouldn't be much legal risk until the agreements get much larger.
Cheers,
James
There's a lot here, I like these lines of thinking.
> This incentive to cooperate is one way that an accuracy agreement could be a better structure than a prediction market.
I could see this, though if actors can resell their contracts, then there might be some privacy. A potential buyer of a contract might underplay their hand, but actually, know of clever ways of reducing the uncertainty more than expected.
Meanwhile in prediction markets, there are of course actors who buy a big trade, then share selective information to support their trade. So there's at least some collaboration in each.
I'm most excited about collaboration via additional services of just estimating the value of collaboration and directly rewarding people, but that's more complicated.
>Actually, according to your example, bidders submit their forecasts with the bid. Seems like because of research and analysis costs they should only submit forecasts after winning the bid
I think "submitting forecasts with the bid" is simpler, for a thought experiment, but in practice mediocre, for this reason. I'd be more optimistic about submitting forecasts after, or being "continuously scored" using updating forecasts.
For example, a certain party could have been paid to forecast COVID cases, then be expected to forecast weekly - then later paid conditional on their accuracy.
> the purchaser could submit qualitative predictions (e.g. prose) and the client could later give a score for how well they did.
This is a bit of a different topic, but I've thought about this, and could be excited by it, especially if we could use AI. Like, we could hypothetically use AI to judge the accuracy of qualitative forecasts made by all historical figures and twitter intellectuals.
> An unsolved problem for prediction markets with an AMM is news traders taking more profit relative to effort expended. This accuracy agreement doesn't solve this problem.
I imagine the ease would be "priced in". If predictors thought that the first 3 bits of information are very easy to get, they'd submit an according bid for that.
>If a decisive event happens early that could earn them a lot of money, then they could rest and vest.
Sure, but they still could make even more money by doing better. They could also sell the contract/agreement, to someone interested in extracting more money from it.
>In fact, because of the potential for profit, there could be some incentive for them to act to cause certain outcomes.
True! This is a problem.